Multi-path Routing for Supercomputing User Experience
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-path routing methods, such as ECMP, do not effectively consider differences in network features and data stream requirements, leading to suboptimal performance when path differences are significant or data stream requirements vary.
Innovation Solution
A multi-path routing method that decouples services into blocks, acquires network requirement features, and uses a matching degree evaluation function to select paths based on network features, improving user experience by formally describing multi-dimensional and fine-grained requirements of supercomputing applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If ECMP routing is used, then routing simplicity is maintained, but routing effectiveness deteriorates due to ignoring network feature differences and data stream requirements
Solution Approach 1:
The service is decoupled into multiple service blocks according to a preset rule, and each service block is independently matched with appropriate network paths based on its specific network requirement features. This segmentation allows different parts of the service to utilize different network paths optimized for their specific requirements, resolving the contradiction between routing simplicity and routing effectiveness.
Solution Approach 2:
The patent introduces a matching degree evaluation function that takes network features and service requirement features as parameters to dynamically determine the appropriate path for each service block. By changing the routing parameters from simple equal-cost selection to feature-based matching, the system achieves both operational simplicity through automation and improved routing effectiveness.
2Reliability
If service decoupling and feature-based matching are implemented, then routing effectiveness is improved, but system complexity increases
Solution Approach 1:
The system automatically performs service decoupling, feature extraction, path candidate generation, and matching degree evaluation without requiring manual intervention. The preset rule and matching degree evaluation function enable the system to self-configure optimal paths based on network conditions and service requirements, improving routing effectiveness while managing complexity through automation rather than manual configuration.
Solution Approach 2:
The matching degree evaluation function serves multiple purposes: it evaluates path suitability for different service blocks, incorporates various network features (bandwidth, delay, reliability), and adapts to different service types. This multi-functional approach consolidates multiple routing decisions into a single unified mechanism, improving routing effectiveness without proportionally increasing system complexity.
3Speed
If path selection ignores network features, then processing speed is maintained, but user experience quality deteriorates
Solution Approach 1:
The system pre-calculates and stores network features of different paths and service requirement features of service blocks before actual routing decisions are needed. The matching degree evaluation function can quickly retrieve and compare these pre-prepared features, maintaining processing speed while enabling feature-based path selection that improves user experience quality.
Solution Approach 2:
The patent replaces manual or simple mechanical path selection with an automated evaluation function that systematically compares network features with service requirements. This substitution of automated computational processes for manual or simplistic routing maintains high processing speed while dramatically improving routing quality and user experience through feature-based decision making.
Data Source
AI summary
A multi-path routing method oriented to supercomputing user experience quality includes: decoupling, according to a preset rule, a service with a path to be planned into at least one service block; acquiring, according to a network requirement feature of each service block, all paths between network nodes of the path to be planned, and a network feature of each of all the paths, a multi-path set between the network nodes for the service; and inputting the network feature of each path in the multi-path set and the network requirement features of all the service blocks into a preset matching degree evaluation function to acquire a network path between the network nodes for the service. According to the multi-path routing method, the overall service of the network is described in blocks, so that a strong dependency relationship between supercomputing service task scheduling and data exchange is decoupled.